Dissecting Behaviour: Relations Between Autoaggression, Grooming and Walking in a Macaque

Publisher Website
Google Scholar
Abstract
When considering the alternation of different activities shown during a certain observation period from a causal point of view, many ethologists interpret the observed behaviour as the outcome of an interplay between a number of tendencies to display the various activities. Such hypothetical underlying behaviour mechanisms, however, have often resisted formal quantitative analysis. The reason is that the usual quantitative activity measures such as total time or frequency are not well suited to such a causal analysis when applied to activities which exclude each other in time. These measures retain little or no information about the time structure of the observed behaviour and, therefore, cannot provide a basis for inferences concerning underlying quantitative dependencies between the occurrence of the respective activities. If, as often happens, the behaviour under consideration is highly variable in its time structure, there is a possibility that this structure can be represented by a continuous time Markov chain. If such a representation proves to be legitimate, the aforementioned behavioural tendencies are adequately, quantitatively represented by the state transition rates of the model. These transition rates then contain in principle all information about bout duration as well as sequential dependence (and hence also frequency) of the activities (states). The values for each of the transition rates are not constrained by the values for the others. Moreover, their estimators are (asymptotically) statistically independent. This permits a full analysis of the interplay of the respective tendencies within an observation period as well as in the variation between observation periods. The present analysis is based on nine records of 15 minutes each of the behaviour of a singly caged stumptailed macaque. It appeared that the behaviour structure could adequately be described as a Markov chain with six states: walking, scratching, autogrooming, autoaggression and two types of blank intervals (during which the animal was doing nothing much in particular except for sitting and looking around). The two types ofblanks were distinguished on the basis of their average duration and the type of activities that followed. The analysis is given in the first part of the paper. This part also illustrates that particular deviations from Markovity tend to occur if activities do not correspond to only one underlying Markovian state, e.g., due to an unlucky choice of behavioural categories. The Markovian representations of all records together showed that autoaggression and grooming tendencies were closely linked; both were especially high during blank 2 and scratch. However, on the average the occurrence of grooming led to a considerable decrease in the tendencies for aggression as well as other activities. The second part of the paper presents an analysis of the variation in transition rates across the records. Particular combinations of levels of tendencies and of correlations between tendencies were interpreted in terms of hypothetical mechanisms governing their interplay. The subject's overall behaviour appeared to be strongly governed by a variable degree of " restlessness", expressing itself in negative correlations of the tendencies for replacing walking for any of the two types of blanks with the tendency for resuming walking during one type of blank. In this interpretation, the restlessness seemed to reduce the amount of autoaggression in three different ways: by "postponement" of the onset of aggression, by "suppression" once it had begun and by "prevention" of aggression after scratch because of intervening resumption of walking. With regard to grooming the results were suggestive of an aggression decreasing effect in two different ways: by "replacement" of aggression after a blank 2 interval and by "forestalling" as well as postponing aggression after a bout of scratching. When considering the alternation of different activities shown during a certain observation period from a causal point of view, many ethologists interpret the observed behaviour as the outcome of an interplay between a number of tendencies to display the various activities. Such hypothetical underlying behaviour mechanisms, however, have often resisted formal quantitative analysis. The reason is that the usual quantitative activity measures such as total time or frequency are not well suited to such a causal analysis when applied to activities which exclude each other in time. These measures retain little or no information about the time structure of the observed behaviour and, therefore, cannot provide a basis for inferences concerning underlying quantitative dependencies between the occurrence of the respective activities. If, as often happens, the behaviour under consideration is highly variable in its time structure, there is a possibility that this structure can be represented by a continuous time Markov chain. If such a representation proves to be legitimate, the aforementioned behavioural tendencies are adequately, quantitatively represented by the state transition rates of the model. These transition rates then contain in principle all information about bout duration as well as sequential dependence (and hence also frequency) of the activities (states). The values for each of the transition rates are not constrained by the values for the others. Moreover, their estimators are (asymptotically) statistically independent. This permits a full analysis of the interplay of the respective tendencies within an observation period as well as in the variation between observation periods. The present analysis is based on nine records of 15 minutes each of the behaviour of a singly caged stumptailed macaque. It appeared that the behaviour structure could adequately be described as a Markov chain with six states: walking, scratching, autogrooming, autoaggression and two types of blank intervals (during which the animal was...

This publication has 2 references indexed in Scilit: